智能工作流 是 AI Skill Hub 本期精选Agent工作流之一。综合评分 8.0 分,整体质量较高。我们强烈推荐将其纳入你的 AI 工具库,帮助提升工作效率。
智能工作流 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
智能工作流 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排,将复杂的多步骤任务拆解为清晰的自动化流程,实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成,适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一:cargo install(推荐) cargo install ilmari # 方式二:从源码编译 git clone https://github.com/bnomei/ilmari cd ilmari cargo build --release # 二进制在 ./target/release/ilmari
# 查看帮助 ilmari --help # 基本运行 ilmari [options] <input> # 详细使用说明请查阅文档 # https://github.com/bnomei/ilmari
# ilmari 配置说明 # 查看配置选项 ilmari --config-example > config.yml # 常见配置项 # output_dir: ./output # log_level: info # workers: 4 # 环境变量(覆盖配置文件) export ILMARI_CONFIG="/path/to/config.yml"
Ilmari is a tmux popup radar for coding-agent panes.
It scans the tmux panes you already have running, detects supported agent CLIs, groups panes by workspace, shows each pane's state and recent output, and jumps you back to the selected pane. It is observer-only: it does not launch agents, manage workflows, or own your tmux layout.
tmux is installed and running.cargo install ilmari
Verify the binary is on your PATH:
ilmari --version
Expected output:
ilmari <version>
Download the matching prebuilt release binary instead of compiling it locally:
cargo binstall ilmari
The default Cargo build includes the TUI and both publishing endpoints:
cargo build
You can remove features when building:
cargo build --no-default-features
cargo build --no-default-features --features tui
cargo build --no-default-features --features socket
cargo build --no-default-features --features mcp
cargo build --no-default-features --features rmcp
cargo build --no-default-features --features socket,mcp
cargo build --no-default-features --features tui,socket,mcp
| Feature | Effect |
|---|---|
tui | Builds the terminal UI and pulls in ratatui and crossterm. |
socket | Builds the local Unix JSON socket endpoint. |
mcp | Builds the loopback MCP resource endpoint and pulls in rmcp, axum, tokio, and tokio-util. |
rmcp | Alias for mcp, for builds that name the backing crate explicitly. |
Runtime flags remain parseable when an endpoint is compiled out. Enabling a compiled-out endpoint reports a status warning such as `socket support was not compiled in or MCP support was not compiled in. Builds without feature tui` run in headless mode.
Cause: the binary was built without the socket feature.
Fix:
cargo build --features socket
Cause: the binary was built without the mcp feature.
Fix:
cargo build --features mcp
No configuration file is required. When present, Ilmari reads $XDG_CONFIG_HOME/ilmari/config.toml, falling back to ~/.config/ilmari/config.toml. Malformed TOML, an unknown table/field, or an invalid typed value produces a clear startup error rather than being silently ignored.
This complete example also shows the built-in defaults:
```toml [runtime] refresh_seconds = 5 process_refresh_seconds = 15
[scanner] git = true output_tail = true
[tui] enabled = true bell = true
[palette]
@ilmari_status_summary is the recommended global fragment. For a custom tmux format or menu helper, notification totals remain available as @ilmari_waiting_count and @ilmari_finished_count, while @ilmari_running_count is the live running count and @ilmari_attention_count is the combined sticky total. Ordinary counts are published separately as @ilmari_waiting_state_count, @ilmari_finished_state_count, @ilmari_terminated_count, and @ilmari_unknown_count, so no pane is counted in both attention and an ordinary state. Pane-local @ilmari_state and @ilmari_badge are available to advanced custom formats. Pane-local @ilmari_attention is 0 or 1 and carries the sticky attention latch across a popup's failed daemon snapshot followed by a direct scan; it is cleared with the other pane options when state is cleaned up or when the focused pane acknowledges attention.
Running state is live. Waiting-input and finished attention is sticky only after a qualifying transition occurs while that exact pane is not focused by any tmux client. Focusing the pane acknowledges it. An unchanged later scan does not recreate attention, but a later qualifying transition can. State for disappeared panes is removed.
Before uninstalling the TPM plugin, either set @ilmari_daemon off and reload tmux or run ilmari daemon stop from that tmux server. This stops its daemon and clears Ilmari's published fragments, counts, pane state, socket path, and MCP URL. Then remove the @plugin line and uninstall through TPM.
高质量的开源AI工作流项目
AI Skill Hub 为第三方内容聚合平台,本页面信息基于公开数据整理,不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后,再部署至生产环境,并做好必要的安全评估。
✅ MIT 协议 — 最宽松的开源协议之一,可自由商用、修改、分发,仅需保留版权声明。
经综合评估,智能工作流 在Agent工作流赛道中表现稳健,质量优秀。如果你已有明确的使用需求,可以直接上手体验;如果还在评估阶段,建议对比同类工具后再做决策。
| 原始名称 | ilmari |
| 原始描述 | 开源AI工作流:Minimal tmux popup radar for Codex, Amp, Claude Code, OpenCode, Pi, Gemini, Augg。⭐15 · Rust |
| Topics | AIRust工作流 |
| GitHub | https://github.com/bnomei/ilmari |
| License | MIT |
| 语言 | Rust |
收录时间:2026-06-17 · 更新时间:2026-06-20 · License:MIT · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型,复制安装指令后粘贴到对应客户端